Early 90nm Athlon64s hitting 3GHz?

[clarkey01]

90 I think, FX 55 will be on 130 nm, not bad for 2.6Ghz hay.

 

[jiffylube1024]

Originally posted by: Viditor

Originally posted by: AWhackWhiteBoy
it scales linearly, its barely behind the AMD64 lines mhz for mhz, and its new and has LOTS of room to scale. it also uses very,very little power so putting two in one core could make an amazing, average power consuming processor.

Dothan's power usage number is very misleading...

1. When you measure power usage on a Dothan vs Mobile A64, don't forget that you must add the 6-9 watts of the northbridge to the Dothan's power (A64 has it included in the CPU)
2. Dothan's power is measured as an average, not max...A64 is measured at theoretical max.

It doesn't change the fact that the Pentium-M has much better battery life than competing chips.

Originally posted by: infestedgh0st
donathans cannot be underestimated......it's the secrect weapon of intel, at least in 90 nm technology

however, hitting 3 ghz on a64 plateform will just allow amd to leap forward one more step, which makes it even harder for intel to keep up
amd with SOI should hit 3 ghz without much trouble on heat, however, if amd64 gets even more efficient at 90nm compare to 130nm, then intel is going hav to play a big time catch-up

the dual core is teh ownage, i've read about amd's "double gate" cpu almost a year ago....i guess the day has finally come

Intel has said repeatedly that Prescott is what they are sticking with in the desktop market for the future, even with dualcore.

So, while on paper, when you get your noodle going ("if they could only transplant Dothan to a desktop chip, maybe in dualcore form, run some more current through it and o/c the crap out of it, essentially, with Desktop PC's superior cooling, then they'd really have something).

However, for the moment at least, it looks like Prescott is Intel's desktop flagship. Dothan is designed for the mobile market, and it would take some tweaking and clever thinking to allow it to flourish into the desktop beast that it *seems* it could be. Remember, this is all in theory right now though. Nobody truly knows if a hard switch to dualcore Dothan would magically make Intel the market leader again.


Right now I'm pretty annoyed at how stagnant it is in the (Intel) CPU market right now. Prescott, DDR-2, etc is reminding me of the Pentium 4 launch; evolutionary technology that isn't faster than the old, at least not yet.

Just like the original P4 wasn't any faster than the P3 (and especially when the 0.13um shrink happend on the P3 later), Prescott isn't impressing much over Northwood yet. It will improve eventually though (or else Intel's f*cked ).

AMD is looking to be really in the driver's seat right now, though, and if they can get both 90nm and dualcore out within the next year, well then it might be time to buy some AMD stock . Right now, Hyperthreading is the only reason I'm running Intel!

 

[clarkey01]

jiffylube1024 there goona beef up the presscot with a 1066 Mhz FSB and another 1 meg of cache (2mb), what this will translate into performance ? I dont know.

 

[imported_SLIM]

Originally posted by: AdamUK
What nm is the AMD 64 4000? And when will that be out?

This article at the inq says that the 4000 and fx55 are going to be 130nm parts (probably the last speed bump for that process).
http://www.theinquirer.net/?article=18540

SLIM

 

[Jeff7181]

Originally posted by: Marlin1975

Originally posted by: AWhackWhiteBoy<BR>i hope no one forgot about the sleeping giant called Dothan.

<BR><BR><BR><BR>What about it. It will only get to 2.1Ghz by years end and a 2Ghz version is around the same as a 3Ghz P4.

Dude... you work in Intel's R&amp;D labs and know this for a fact? Why didn't you tell us this sooner? :roll:

 

[clarkey01]

The most important thing I?d like to get across is how the A64 scales and the numbering system + Prescott stuff.

I myself criticized the 4000+ ?s name, running @ only 2.4Ghz I was kind of shocked but you have to see the reasoning behind it.

A) Either they cant get to 2.6Ghz on 90nm ( Very unlikely, unless you?re an Ed fan over @ overclockers.com).

B) They are holding back because they wont have 65 nm capability until FAB 36 comes online and yields are acceptable. So they have to hold back how much they scale, I myself believe the A64 will see 3ghz and maybe a little over but they have to have some Mhz in reserve, so the 4200+ will be running @ 2.6Gh.



Intel fans and some AMD fans may go ?4200+ 2.6ghz doesn?t stand up to its name?, Remember a 200Mhz boost in a K8 CPU is more then a 200Mhz jump in a Netburst CPU, not to mention the 90 nm improvements as shown below and posted in previous threads:

Full SSE3 implementation
* Improved hardware data prefetch mechanism
* Increased number of writing combine buffers (D0 stepping A64's can now combine up to four non-cacheable streams compared to 2 on the C0 and CG stepping A64's)
* Improved on-die memory controller with more advanced open page policy
* On-die thermal throttling
* Black Diamond Low-K technology (slower less power hungry transistors in less used sections and faster and more power hungry transistors in frequently used sections of the cpu)

Getting Prescott up to 4.2Ghz with a 1066 Mhz FSB and 2 Meg of Cache, I mean, what are the yields going to be like ? I agree Intel has the leading manufacturing process sin the world (CPU) but Prescott is a dead horse which they are struggling to get enough 3.4?s/3.6?s and have the released a 3.8Ghz ?

They seem hard up as it with 800 FSB, 1Meg Cache, 3.6ghz cloclspeed, I doubt an extra 800 Mhz, 1066Mhz bus, another 1 meg of cache will help. Yields are hard to predict but I have a bad feeling they wont be better then current 3.4?s,3.6?s.

 

[Drayvn]

I dont know about the specs for the +4000

My guessing is it will be much like th FX-55 just without some of the features of the FX-55, making it much like the +3800, which is much like the same as the FX-53, but without some of its features.

 

[LocutusX]

Originally posted by: Megatomic
With only a few short exceptions I've been all AMD all the time since my AMD 486-33. It hasn't always been a fast ride but it has been an interesting one.

I was an Intel boy through the Coppermine era and only jumped ship once the Palomino came out. Sure, it wasn't their greatest chip but I haven't regretted it.

Funnily enough, all my CPUs have been "special" in some way...

1. Pentium Pro 166mhz - P6 architecture still in use today (Pentium-M), Integrated Copper Heatspreader (!), 512kb on-die cache running at full speed... (sounds like a modern-day CPU)
2. Coppermine 600E @ 850mhz... 42% overclock, 'nuff said
3. Palomino @ 1.5GHz... poor overclocker, but was a lot of fun. finished the last 3 years of games incl. Doom3 on it. when i got this, it was apparent AMD was going to take the lead someday.
4. Newcastle @ 2.4GHz... 64-bit, finally! and back to the integrated copper heatspreader from 1997

 

[Uhtrinity]

Originally posted by: LocutusX

Originally posted by: Megatomic
With only a few short exceptions I've been all AMD all the time since my AMD 486-33. It hasn't always been a fast ride but it has been an interesting one.

1. Pentium Pro 166mhz - P6 architecture still in use today (Pentium-M), Integrated Copper Heatspreader (!), 512kb on-die cache running at full speed... (sounds like a modern-day CPU)

Not to be too picky, but the Pentium pros had l2 Cache on chip, not on die. Theye actually mounted two dies into one ceramic case. One die was the CPU and the second was the cache.

 

[LearnedPadawan]

It seems AMD is catching up when it comes to pure speed as well. Intel = Bankrupt in a decade.

 

[jm0ris0n]

Hmmm...Summer 2005....think we'll have 90nm A64s that'll OC to 3Ghz+ on air (high-end air) with chips themselves running under $150...?

Hmmm, my summer 2005 wish list:

Magical 3Ghz+ (on OC) A64 CPU, 150
Dual PCI Express NForce, 200

= YaY ! :-D

 

[Zebo]

Originally posted by: LearnedPadawan
It seems AMD is catching up when it comes to pure speed as well. Intel = Bankrupt in a decade.

LOL

Intels in a bit of a pickle for sure, if the 90nm chips run cool, scale to 3.2 within the next year, Intel is in a world of hurt. But only with us, the enthusiasts. Mom and pop and susie Q will always buy Intel.

 

[Viditor]

It doesn't change the fact that the Pentium-M has much better battery life than competing chips

I believe you are wrong in that...

1. Intel's TDP for the 2.4 GHz Pentium-M is 30W...add the northbridge and that's 36-39W as an average useage
2. AMD's mobile A64s use 35W as a theoretical Max...

 

[Viditor]

Originally posted by: AdamUK
What nm is the AMD 64 4000? And when will that be out?

130nm when it's first released (IIRC), subject to change to 90nm as the FX53 inventories go away

 

[rgreen83]

Originally posted by: Viditor

It doesn't change the fact that the Pentium-M has much better battery life than competing chips

I believe you are wrong in that...

1. Intel's TDP for the 2.4 GHz Pentium-M is 30W...add the northbridge and that's 36-39W as an average useage
2. AMD's mobile A64s use 35W as a theoretical Max...

Pentium-Ms are already to 2.4ghz? I thought they were still at 1.8 or 2.0? I am a big fan of the athlon 64s (have one myself now), but the people saying that intel should be worried are taking it a bit far, they are 10x more broadly based than amd, even if they start to lose the cpu race, they will just move to something else. They have more fabs and manufacturing capabilites than any other company in the world, heck if they lost the race, amd would have to hire them to make athlons because amd cant come close to satisfying the market. And dont forget, intel is currently the #1 seller of mainboards, chipsets, and graphics in addition to cpus.

:beer: heres to competition at its finest, may it bring us ever better and cheaper stuff! :beer:

 

[jiffylube1024]

Originally posted by: clarkey01



Getting Prescott up to 4.2Ghz with a 1066 Mhz FSB and 2 Meg of Cache, I mean, what are the yields going to be like ? I agree Intel has the leading manufacturing process sin the world (CPU) but Prescott is a dead horse which they are struggling to get enough 3.4?s/3.6?s and have the released a 3.8Ghz ?

They seem hard up as it with 800 FSB, 1Meg Cache, 3.6ghz cloclspeed, I doubt an extra 800 Mhz, 1066Mhz bus, another 1 meg of cache will help. Yields are hard to predict but I have a bad feeling they wont be better then current 3.4?s,3.6?s.

Going to 1066MHz FSB isn't an issue - you can o/c an unlocked P4 to this easily already, and this was possible with lower Northwoods as well (like newer 2.4 C's). Higher FSB doesn't really affect the CPU.

2MB of cache isn't really a yield issue either, it just means more transistors on die and thus less profit for Intel as they get fewer chips per wafer (but a necessary step to counter AMD's superior performance).

The only "issue" is getting Prescott to 4.2 Ghz, and that will be the biggie based on how bad Presshott has been scaling thus far.

 

[rgreen83]

Originally posted by: jiffylube1024

Originally posted by: clarkey01



Getting Prescott up to 4.2Ghz with a 1066 Mhz FSB and 2 Meg of Cache, I mean, what are the yields going to be like ? I agree Intel has the leading manufacturing process sin the world (CPU) but Prescott is a dead horse which they are struggling to get enough 3.4?s/3.6?s and have the released a 3.8Ghz ?

They seem hard up as it with 800 FSB, 1Meg Cache, 3.6ghz cloclspeed, I doubt an extra 800 Mhz, 1066Mhz bus, another 1 meg of cache will help. Yields are hard to predict but I have a bad feeling they wont be better then current 3.4?s,3.6?s.

Going to 1066MHz FSB isn't an issue - you can o/c an unlocked P4 to this easily already, and this was possible with lower Northwoods as well (like newer 2.4 C's). Higher FSB doesn't really affect the CPU.

2MB of cache isn't really a yield issue either, it just means more transistors on die and thus less profit for Intel as they get fewer chips per wafer (but a necessary step to counter AMD's superior performance).

The only "issue" is getting Prescott to 4.2 Ghz, and that will be the biggie based on how bad Presshott has been scaling thus far.

Wouldnt the extra cache add a lot of extra transistors that would produce more heat? making even more difficult for them to get to 4 or 4.2 ghz?

 

[Accord99]

Originally posted by: Viditor

It doesn't change the fact that the Pentium-M has much better battery life than competing chips

1. Intel's TDP for the 2.4 GHz Pentium-M is 30W...add the northbridge and that's 36-39W as an average useage

It's 21W for the 2GHz Dothan. Intel's TDP is determined using a variety of CPU intensive applications and set at a level above the maximum sustained power dissipation. More important for laptops is low CPU usage and idle power dissipation, in which the Pentium-M is significantly better.

 

[clarkey01]

"Wouldnt the extra cache add a lot of extra transistors that would produce more heat? making even more difficult for them to get to 4 or 4.2 ghz? "

rgreen83 your right, More cache means more transitors , which means more leakage, which means more heat.


"
Going to 1066MHz FSB isn't an issue - you can o/c an unlocked P4 to this easily already, and this was possible with lower Northwoods as well (like newer 2.4 C's). Higher FSB doesn't really affect the CPU.

2MB of cache isn't really a yield issue either, it just means more transistors on die and thus less profit for Intel as they get fewer chips per wafer (but a necessary step to counter AMD's superior performance).

The only "issue" is getting Prescott to 4.2 Ghz, and that will be the biggie based on how bad Presshott has been scaling thus far."

I know that getting the FSB up to 1066 costs Intel, dont ask me how but I kno wand I'll try and link you, the more cache you have on a chip the less likely its going to pull off, why do you think AMD cut the clawhammers 1 Meg in half ? yes it does reduce die space and also allows them to sell 512k ones know as "Newcastles" which otherwise would have been dead CPU'S. As for hitting 4.2Ghz, I doubt it very much.

 

[Zebo]

I wonder if the Nforce 4 will be 250FSB then AMD will even steal the 1066 thunder.

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor

It doesn't change the fact that the Pentium-M has much better battery life than competing chips

1. Intel's TDP for the 2.4 GHz Pentium-M is 30W...add the northbridge and that's 36-39W as an average useage

It's 21W for the 2GHz Dothan. Intel's TDP is determined using a variety of CPU intensive applications and set at a level above the maximum sustained power dissipation. More important for laptops is low CPU usage and idle power dissipation, in which the Pentium-M is significantly better.

From the Intel spec sheet for TDP...

?The numbers in this column reflect Intel?s recommended design point and are not indicative of the maximum power the processor can dissipate under worst case conditions.?

From AMD's spec sheet on TDP...

Thermal Design Power (TDP) is measured under the conditions of TCASE Max, IDD Max, and VDD=VID_VDD, and include all power dissipated on-die from VDD, VDDIO, VLDT, VTT, and VDDA.?

These 2 measurements are quite different...
I must admit that I extrapolated the 30W for the 2.4 GHz Dothan from overclocked data (oops...), but nonetheless, the Athlon64 mobile line will not increase it's TDP as clockspeed increases (AMD's numbers are calculated for the entire line to max calculated speed), whereas the Intel TDP can increase with each stepping. Also, the 21W you quoted was not the maximum TDP, and did not include the power for the Northbridge or memory controller (which the A64 did).
I do doft my hat to Intel marketing, but I don't really think that they are necessarily lower power chips (in fact, I believe the opposite is true...)

 

[Accord99]

Originally posted by: Viditor
From the Intel spec sheet for TDP...

?The numbers in this column reflect Intel?s recommended design point and are not indicative of the maximum power the processor can dissipate under worst case conditions.?

Intel's methodolgy for TDP determination comes from here:
ftp://download.intel.com/desig...m4/guides/24920301.pdf (page 25)
Unless you run a power virus, you will not exceed the TDP.

whereas the Intel TDP can increase with each stepping.

Not for the Pentium-M so far. Each class of Pentium-Ms has shared a common TDP.

I do doft my hat to Intel marketing, but I don't really think that they are necessarily lower power chips (in fact, I believe the opposite is true...)

None of the notebook reviews with battery testing support your theory, Pentium-M based laptops have an edge against similarly equipped A64 at high CPU load situations and have a huge advantage, as much as 50%, in low CPU usage after normalizing for battery size. Even against 25W LV AXP-Ms, the 24.5W TDP P-Ms had better performance and battery life.

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
From the Intel spec sheet for TDP...

?The numbers in this column reflect Intel?s recommended design point and are not indicative of the maximum power the processor can dissipate under worst case conditions.?

Intel's methodolgy for TDP determination comes from here:
<a target=_blank class=ftalternatingbarlinklarge href="ftp://download.intel.com/design/Pentium4/guides/24920301.pdf">ftp://download.intel.com/de......uides/24920301.pdf</a> (page 25)

Interesting (I wonder if they've changed since 2000 when this was written...?). It appears that the TDP for Intel is set at 75% of Max power of their proprietary software (though that "Max" is lower than theoretical max). So to compare apples to apples, we get 21.5W*4/3=26.7W...add the memory controller and Northbridge (10W for the 82875P), and we get a generous (I say generous because it's still not peak theoretical) TDP of 36.7W as compared to the A64's 35W at peak (actual power on the A64 mobile will probably be closer to 30W at 2GHz)

Unless you run a power virus, you will not exceed the TDP.

Pardon me if I tend to doubt that...I imagine that even playing UT2004 at higher res will exceed their software suite.

whereas the Intel TDP can increase with each stepping.

Not for the Pentium-M so far. Each class of Pentium-Ms has shared a common TDP.

You are quite right, and I thank you for the correction! This has not been true for any of the other Intel chips, so I made the mistake of "assuming" (I hate it when I do that!!).

I do doft my hat to Intel marketing, but I don't really think that they are necessarily lower power chips (in fact, I believe the opposite is true...)

None of the notebook reviews with battery testing support your theory, Pentium-M based laptops have an edge against similarly equipped A64 at high CPU load situations and have a huge advantage, as much as 50%, in low CPU usage after normalizing for battery size. Even against 25W LV AXP-Ms, the 24.5W TDP P-Ms had better performance and battery life.
[/quote]

I believe this has much more to do with the laptop design than the CPU...it's not really correct to attribute it to the P-M (IMHO)...

 

[Accord99]

Originally posted by: Viditor
Interesting (I wonder if they've changed since 2000 when this was written...?). It appears that the TDP for Intel is set at 75% of Max power of their proprietary software (though that "Max" is lower than theoretical max).

Northwood had a value of 85%. Banias and Dothan are probably even closer to their max, and this max is a theoretical max.

So to compare apples to apples, we get 21.5W*4/3=26.7W...add the memory controller and Northbridge (10W for the 82875P),

Intel's datasheet for the 855PM chipset gives a value of 1.8W.

Pardon me if I tend to doubt that...I imagine that even playing UT2004 at higher res will exceed their software suite.

Games aren't particularly CPU intensive, a lot of time is spent at the video card so they cause the CPU to produce as much heat as the CPU intensive applications used in the testing.

I believe this has much more to do with the laptop design than the CPU...it's not really correct to attribute it to the P-M (IMHO)...

And when laptops have similar components and screens, then the only variables are the CPU and the chipset. And the proof is in the superior battery life in reviews and the wide availabilty of small laptops using the P-M versus none for the A64.

 

[rgreen83]

And when laptops have similar components and screens, then the only variables are the CPU and the chipset. And the proof is in the superior battery life in reviews and the wide availabilty of small laptops using the P-M versus none for the A64.

I think it does have some to do with marketing, A64 are marketed as high performance, and are thus put in high performance machines, the opposite is true for the Pentium-M, even though it has shown to be very powerful it is generally marketed and sold with a conservationist approach ( hence no higher end,power-hungry graphics for them).